Submitted to: ASA-CSSA-SSSA Annual Meeting Abstracts
Publication Type: Abstract Only
Publication Acceptance Date: 11/4/2004
Publication Date: 11/4/2004
Citation: Meek, D.W. 2004. Recommendations for Fitting Log Profile Data [CD-ROM]. ASA-CSSA-SSSA Annual Meeting Abstracts. Madison, Wisconsin.
Technical Abstract: Various graphical and analytical techniques exist for estimating the parameters from log profile data. Are they all adequate and equal? This study evaluates three different regression-based methods; they are unweighted least-squares, weighted least-squares, and an integral log profile model via unweighted least-squares (a.k.a., mass conservation). Multiple data sets for neutral conditions were selected from several well-known publications. Multiple regression performance measures as well as diagnostics were considered. Unweighted least-squares estimation, the most commonly used method, is not recommended because the error structure is not homoscedastic and often has other problems. Weighted least-squares estimation performs much better and results in better assessment of the parameter uncertainty. The selection of an appropriate weight, however, generally requires extra effort and additional computing for each data set. Mass conservation preformed seemingly better than those for the weighted least-squares and generally required just one pass at the dat, but the model needed an offset to adjust for the data integration starting at the displacement height instead of at ground level. Moreover, the latter displacement height estimates are generally larger than those given by the other two methods. If correct, these results suggest that parameter estimates from most other studies could be seriously biased.